context-course/pr_2/en/unit0/introduction.md

# Welcome to The Context Course

This course is about **context engineering** for AI code agents: structuring knowledge so an agent can find what it needs, when it needs it.

## The Key Skill for Code Agents

Claude Code, Codex, and OpenCode all share the same constraint: an agent is only as good as the context it has. Good context means fewer wrong turns, cleaner diffs, and less rework.

Across six units you'll build portable skills, wire up tools through the Model Context Protocol (MCP), package those pieces into plugins, coordinate sub-agents for larger tasks, and study how a minimal agent loop actually works under the hood.

## What You'll Learn

This course is structured into 6 core units:

| Unit | Topic | What You'll Learn |
|------|-------|-------------------|
| **Unit 0** | Onboarding | Course overview, tool setup, prerequisites |
| **Unit 1** | Agent Skills | What skills are, how to build and share them, how agents load them |
| **Unit 2** | Model Context Protocol | MCPs explained, connecting tools and APIs to agents |
| **Unit 3** | Plugins | Building plugins, designing agent workflows |
| **Unit 4** | Sub-agents | Spawning specialized agents, multi-agent patterns |
| **Unit 5** | Bonus: Nano Harness | Advanced optimization, recursive context patterns |

![Course overview showing the 6 units and how they connect](https://huggingface.co/datasets/mcp-course/images/resolve/main/unit0/course-map.svg)

## Prerequisites

Before starting, you should be comfortable with Python basics (variables, functions, loops, and file I/O), able to navigate directories and run scripts from the command line, and have a Hugging Face account ([huggingface.co](https://huggingface.co)). You'll also need at least one code agent installed and configured — see the setup section below.

## Tool Setup

This course works with multiple code agents. Choose at least one to follow along:

Claude Code is Anthropic's official code agent, accessible via the web, desktop app, or CLI.

```bash
curl -fsSL https://claude.ai/install.sh | bash
```

**Getting started**: Visit [claude.ai/code](https://claude.ai/code) to use Claude Code on the web, or install the CLI above and run `claude` in any project directory. You'll be prompted to sign in on first use.

Codex is OpenAI's code agent with multi-agent capabilities.

```bash
npm install -g @openai/codex
```

**Getting started**: Run `codex` and select **Sign in with ChatGPT** to authenticate with a paid ChatGPT plan (Plus, Pro, Business, Edu, or Enterprise), or use an OpenAI API key.

OpenCode is an open source code agent from [opencode.ai](https://opencode.ai):

```bash
curl -fsSL https://opencode.ai/install | bash
```

Or via npm:

```bash
npm install -g opencode-ai
```

**Getting started**: Run `opencode` in any project directory. OpenCode supports multiple LLM providers — configure your preferred provider on first launch.

## How to Navigate This Course

### Recommended Pace

Plan on one unit per week, roughly 2–3 hours each. Context engineering is a practice-heavy skill, so build the examples rather than skimming them.

### Learning Format

Each unit mixes conceptual material with runnable code, a hands-on project, and a short quiz.

### Customizing Your Path

While we recommend following units in order, you can customize based on your needs:
- **Just want skills?** Start with Unit 1, revisit MCPs when needed
- **Building a plugin for your team?** Start with Unit 3
- **Multi-agent systems?** Begin with Unit 4, return to Unit 1-2 as reference
- **Following along with open source?** All lessons include OpenCode examples alongside Claude Code and Codex

## Certifications

This course offers two levels of certification:

### Context Fundamentals Certificate

Demonstrates you understand core context engineering concepts. Pass the Unit 1–2 quizzes with 70% or higher to earn this certificate in 2–3 weeks. It's shareable and displayed on your Hugging Face profile.

### Context Engineering Certificate

Demonstrates mastery of context engineering across all domains. Pass all Unit 1–5 quizzes (70% or higher) and complete the capstone project to earn this certificate in 5–8 weeks. It's displayed on your Hugging Face profile with a project showcase.

Both certificates verify your ability to build and maintain agent skills, connect external tools through MCPs, design multi-agent systems, and optimize context for maximum agent performance.

![Certification paths: Fundamentals after Units 1-2, Full Certificate after Units 1-5 plus capstone](https://huggingface.co/datasets/mcp-course/images/resolve/main/unit0/certification-paths.svg)

## Course Structure

Every unit follows the same shape: an introduction, conceptual material, practical walkthroughs, a hands-on project, and a quiz. Starter templates and copy-pasteable code are provided throughout so you spend time on the ideas rather than on boilerplate.

## Meet Your Instructors

**Ben Burtenshaw** — ML Engineer, Hugging Face

Ben focuses on LLM applications with emphasis on post-training and agentic approaches. He leads initiatives around agent best practices and context engineering at Hugging Face.

## Connect with the Community

Learning is better together. Join the conversation:

- **Discord**: [discord.gg/huggingface](https://discord.gg/huggingface)
- **Share your work**: Tag #ContextCourse on social media
- **Report issues**: GitHub Issues for course content bugs

## Next Steps

Install at least one of the agents above, check the prerequisites, then head to Unit 1 to start with agent skills.